perf(species): seed the bundled catalog only once per version

Skip the ~1.5 MB catalog parse and the table scan on every launch — the
main startup cost. `seedBundledIfNeeded` re-seeds only when the recorded
version differs from `speciesCatalogVersion`, writing the version after
the seed commits so an interrupted seed retries next launch.

Adds `parseSpeciesCatalogVersion` and a test keeping the constant in sync
with the asset's `version` field.
This commit is contained in:
vjrj 2026-07-10 15:46:04 +02:00
parent c9c764e624
commit ea5ab0342d
4 changed files with 87 additions and 0 deletions

View file

@ -31,3 +31,16 @@ Future<List<SpeciesSeed>> loadBundledSpecies() async {
final jsonString = await rootBundle.loadString(_catalogAsset);
return parseSpeciesCatalog(jsonString);
}
/// Version of the bundled catalog, mirrored from the asset's `version` field
/// (written by `tool/gen_species_catalog.dart`). A test asserts the two stay in
/// sync. The app records the version it last seeded and skips the ~1.5 MB parse
/// and the table scan in [SpeciesRepository.seedBundled] on every subsequent
/// launch so bumping this (alongside the asset) is what triggers a one-time
/// re-seed after the catalog changes.
const int speciesCatalogVersion = 3;
/// Reads just the `version` field from a catalog JSON string. Used by the sync
/// test to keep [speciesCatalogVersion] aligned with the asset.
int parseSpeciesCatalogVersion(String jsonString) =>
(jsonDecode(jsonString) as Map<String, dynamic>)['version'] as int;

View file

@ -70,6 +70,24 @@ class SpeciesRepository {
/// Reads all existing species once and writes in a single batch: with a
/// catalog of ~1000 species a per-row SELECT on every startup would be a real
/// cost on the encrypted database.
/// Seeds the bundled catalog only when [readVersion] doesn't already report
/// [version]. [loadSeeds] is awaited lazily and skipped entirely when this
/// install already holds [version] so the ~1.5 MB catalog parse and the
/// table scan in [seedBundled] are paid once per catalog version, not on
/// every launch (the main startup cost). On a (re)seed, [writeVersion] records
/// [version] only after [seedBundled] commits, so an interrupted seed is
/// retried on the next launch.
Future<void> seedBundledIfNeeded({
required int version,
required Future<String?> Function() readVersion,
required Future<void> Function(String version) writeVersion,
required Future<List<SpeciesSeed>> Function() loadSeeds,
}) async {
if (await readVersion() == version.toString()) return;
await seedBundled(await loadSeeds());
await writeVersion(version.toString());
}
Future<void> seedBundled(List<SpeciesSeed> seeds) async {
final stamp = Hlc.zero(nodeId).pack();
await _db.transaction(() async {

View file

@ -17,6 +17,16 @@ void main() {
expect(json['version'], 3);
});
test('speciesCatalogVersion mirrors the asset version', () {
// The startup seed guard skips the reseed by comparing this constant to the
// version it last recorded; it must track the asset or a bumped catalog is
// never re-seeded.
expect(
speciesCatalogVersion,
parseSpeciesCatalogVersion(file.readAsStringSync()),
);
});
test('is a broad catalog, not the old starter set', () {
// The expanded catalog holds ~1200 species (was 14 by hand). The floor
// leaves headroom for Wikidata drift across regenerations.

View file

@ -237,4 +237,50 @@ void main() {
// No species named no suggestion.
expect(await repo.classifyLabel('Girasol gigante'), isNull);
});
group('seedBundledIfNeeded', () {
test('seeds on first run and skips the reload once recorded', () async {
String? stored;
var loads = 0;
Future<List<SpeciesSeed>> load() async {
loads++;
return _seeds;
}
await repo.seedBundledIfNeeded(
version: 3,
readVersion: () async => stored,
writeVersion: (v) async => stored = v,
loadSeeds: load,
);
expect(loads, 1);
expect(stored, '3');
expect(await repo.search('Tomato'), isNotEmpty);
// Same version already recorded the ~1.5 MB parse is never loaded again.
await repo.seedBundledIfNeeded(
version: 3,
readVersion: () async => stored,
writeVersion: (v) async => stored = v,
loadSeeds: load,
);
expect(loads, 1);
});
test('reseeds when the catalog version changes', () async {
var stored = '3';
var loads = 0;
await repo.seedBundledIfNeeded(
version: 4,
readVersion: () async => stored,
writeVersion: (v) async => stored = v,
loadSeeds: () async {
loads++;
return _seeds;
},
);
expect(loads, 1);
expect(stored, '4');
});
});
}